Matrix factorization is a popular method for disentangling the mixtures of biological signals that underlie multi-omics data. The resulting lower dimensional representations can be used to infer the extent to which latent processes differ across biological conditions. However, when a multi-omics dataset is generated from only a limited number of samples, the effectiveness of matrix factorization is reduced. Therefore, transfer learning approaches to matrix factorization have previously been proposed and applied to omics data. In this study I simulated multi-omics datasets in order to compare existing and novel transfer learning approaches to matrix factorization. I focused on the Bayesian matrix factorization method MOFA and evaluated approaches with respect to their ability to uncover groundtruth latent structure.